Steam AI disclosure could push AI-made games past 50% by 2027-2028, study says

Steam is about to get crowded with games that have “AI content” in the fine print. And the uncomfortable part is the timeline. In a study shared on his Substack, former Habbo Hotel developer Sulka Haro analyzed approximately 53,600 Steam games launched over a three-year window from July 2023 to July 2026, then ran the trajectory forward. If the current trajectory holds, Haro writes that AI-disclosed games cross roughly 50% of all Steam releases somewhere in 2027-2028.

That prediction is not a vague “someday.” It points to a specific mechanism, too: Steam introduced an AI-disclosure feature in 2024. Haro describes a steady increase in games on Steam admitting to AI-generated content once that field existed. Before 2024, he says there was no Steam page field for developers to disclose their AI usage, so the “baseline” started at literal zero. Against that starting point, nearly 31% of Steam games currently share their AI usage, which Haro characterizes as roughly one in three new Steam games.

Now, here is the executive-friendly question underneath the headline: is this “AI adoption ramp” exponential, or is it more like a slow flood with occasional bursts? Haro’s numbers suggest it is mostly a ramp, not a rocket ship. He reports that nearly 80% of the 37,000 distinct publishers on Steam have never shipped a single AI game. And of the publishers that have shipped AI games, 89% shipped exactly one. Translation: the slop factory is growing, but it is not yet a situation where every publisher is churning out AI content in volume.

Still, Haro argues that the incentives are changing. His premise is that generative AI lowers the barrier not just to make a game, but to make several. That matters because game publishing on Steam is already a distribution machine for volume, ranging from legitimate indies to questionable “shovelware.” Haro adds that for “AI devs,” the shovelware tail (his shorthand for the “5+ bar”) is 3 times higher. In other words, even if most publishers are not using AI yet, those who do may be more likely to produce multiple releases that blend together in discovery.

If you are running a studio, funding a catalog, or sitting on a board that cares about brand and product quality, the next-order implication is discovery, not just creation. Haro’s study lands in the same broader moment where industry analysts are warning about storefront dynamics. Earlier this week, Circana industry analyst Mat Piscatella predicted that game discovery will only get worse on storefronts like Steam as AI “vibe coding” makes it easier for people to clone legitimate games, including indie hits like Peak. Piscatella’s point is that easier production does not automatically translate into better signal. It can translate into more near-duplicates and lower-quality listings that push good games down the feed.

There is also a regulatory and policy angle hiding in the numbers. Steam’s introduction of an AI-disclosure feature in 2024 is essentially a compliance lever: it creates a standardized place to report AI usage. Haro’s claim that the increase begins after the field exists is a reminder that disclosure regimes can change observed behavior as much as underlying tech adoption. When disclosure becomes easy, more developers show up in the dataset, even if they were using AI before. For executives, that means internal metrics about “AI adoption” on a platform need to account for disclosure mechanics, not just engineering decisions.

Finally, consider the reputational and operational risk. Haro’s findings do not say most Steam publishers are turning into full-time AI factories. But they do say AI-disclosed releases are already reaching about one in three, after starting from zero disclosure. If that trajectory continues, decision-makers should treat 2027-2028 as a potential inflection point where “AI in the build” becomes normal enough that customers tune it out, leaving quality differentiation and trust mechanisms to work harder. And in an ecosystem where discovery can degrade, trust and discoverability can become harder to buy, even for teams that do things the right way.

So while the study points to a percentage and a date range, the strategic stakes are broader: how will Steam, publishers, and developers manage the signal-to-noise ratio when AI makes “making several” easier, and when a disclosure field makes it visible? For founders, investors, and operators, the goal is not to panic. It is to plan for a storefront where the baseline content mix shifts, the bargaining power of good curation grows, and “quality” becomes a louder differentiator because the entry cost keeps falling.